serkef commented on a change in pull request #6175: [AIRFLOW-5545] Fixes
recursion in DAG cycle testing
URL: https://github.com/apache/airflow/pull/6175#discussion_r331253041
##########
File path: airflow/models/dag.py
##########
@@ -1467,22 +1467,26 @@ def _test_cycle_helper(self, visit_map, task_id):
"""
from airflow.models.dagbag import DagBag # Avoid circular imports
- # print('Inspecting %s' % task_id)
- if visit_map[task_id] == DagBag.CYCLE_DONE:
- return False
-
- visit_map[task_id] = DagBag.CYCLE_IN_PROGRESS
-
- task = self.task_dict[task_id]
- for descendant_id in task.get_direct_relative_ids():
- if visit_map[descendant_id] == DagBag.CYCLE_IN_PROGRESS:
- msg = "Cycle detected in DAG. Faulty task: {0} to {1}".format(
- task_id, descendant_id)
+ path, visited = [task_id], []
+
+ while path:
+ current_task_id = path.pop()
+ if visit_map[current_task_id] == DagBag.CYCLE_DONE:
+ return False
+ if visit_map[current_task_id] == DagBag.CYCLE_IN_PROGRESS:
+ msg = "Cycle detected in DAG. Faulty task: {}".format(
+ current_task_id)
raise AirflowDagCycleException(msg)
- else:
- self._test_cycle_helper(visit_map, descendant_id)
-
- visit_map[task_id] = DagBag.CYCLE_DONE
+ visit_map[current_task_id] = DagBag.CYCLE_IN_PROGRESS
+ task = self.task_dict[current_task_id]
+ for descendant_id in task.get_direct_relative_ids():
+ path.append(descendant_id)
+ if current_task_id in path:
Review comment:
I'm a bit concerned about this check. `path` is a list and in the context of
this very issue we assume thousands of dags. This check is linear. Do I miss
something? If not, I would suggest benchmarking it against a different
structure (set comes to mind)
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
